ABSTRACT

Much empirical work in the social sciences is devoted to learning individuals' preferences from observing their choice of a product from a finite collection of alternatives (often referred to as "feasible set"). Yet, there is a large body of theoretical and applied literature spanning microeconomics, behavioral economics, marketing, and psychology, suggesting that often individuals do not actually consider every alternative in the feasible set before making their choice. There is also a wide literature documenting that individuals' preferences -- their tastes over different products -- exhibit large heterogeneity even within a group of individuals with similar characteristics. The investigators put forward two broad classes of empirical models of discrete choice that allow both for unobserved heterogeneity in the collection of alternatives that the individual considers, i.e. "consideration set", and in the preferences that each individual holds. In one class of models, heterogeneity in preferences and heterogeneity in consideration sets are allowed to depend on each other. This research develops a method to estimate the distribution of preferences and/or consideration sets, and conduct inference on the estimated distributions. It also develops a method to estimate (and conduct inference on) the welfare effect of policy interventions, e.g. ones that make consumers more aware of specific products or product attributes, or those that change the set of products in the market, etc. A primary output of this research is a collection of portable computer programs implementing the methodology, that will be shared with the community openly and free of charges or restrictions.

This research puts forward new empirical models of discrete choice with unobserved heterogeneity in consideration sets. In the models considered in this research, decision makers are heterogeneous both in the products they consider and in their preferences. The first class of models places no restriction on the consideration set formation process and, in particular, allows for unrestricted forms of dependence of the decision maker's random consideration set with her preferences and with the observable characteristics of the available alternatives. Due to its flexibility, this model is partially but not point identified. The second class of models assumes specific distributions (known up to parameters) for the random consideration sets, building on recent theoretical advances in the microeconomic theory literature on limited consideration/attention. It then aims at providing weak conditions to achieve non-parametric point identification of unobserved heterogeneity in preferences, as well as identification of the distribution of consideration sets. This research aims at suggesting specific estimators/inference methods for the point or set identified distributions. This research further develops computer packages that empirical researcher can use to implement the methods, and that will be shared with the community openly and free of charges or restrictions.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

Please report errors in award information by writing to: awardsearch@nsf.gov.